species distribution modelling for conservation of an

Research Article

Species distribution modelling for conservation of anendangered endemic orchidHsiao-Hsuan Wang1*†, Carissa L. Wonkka2,4†, Michael L. Treglia1,3,5, William E. Grant1,Fred E. Smeins2 and William E. Rogers2

1 Department of Wildlife and Fisheries Sciences, Texas A&M University, College Station, TX 77843, USA2 Department of Ecosystem Science and Management, Texas A&M University, College Station, TX 77843, USA3 Biodiversity Research and Teaching Collections, Applied Biodiversity Science Program, Texas A&M University, College Station,TX 77843, USA4 Present address: Department of Agronomy and Horticulture, University of Nebraska, Lincoln, NE 68583, USA5 Present address: Department of Biological Science, University of Tulsa, Tulsa, OK 74104, USA

Received: 29 August 2014; Accepted: 31 March 2015; Published: 21 April 2015

Associate Editor: Katharine J.M. Dickinson

Citation: Wang H-H, Wonkka CL, Treglia ML, Grant WE, Smeins FE, Rogers WE. 2015. Species distribution modelling for conservation of anendangered endemic orchid. AoB PLANTS 7: plv039; doi:10.1093/aobpla/plv039

Abstract. Concerns regarding the long-term viability of threatened and endangered plant species are increasinglywarranted given the potential impacts of climate change and habitat fragmentation on unstable and isolated popula-tions. Orchidaceae is the largest and most diverse family of flowering plants, but it is currently facing unprecedentedrisks of extinction. Despite substantial conservation emphasis on rare orchids, populations continue to decline. Spir-anthes parksii (Navasota ladies’ tresses) is a federally and state-listed endangered terrestrial orchid endemic to centralTexas. Hence, we aimed to identify potential factors influencing the distribution of the species, quantify the relativeimportance of each factor and determine suitable habitat for future surveys and targeted conservation efforts. We ana-lysed several geo-referenced variables describing climatic conditions and landscape features to identify potential factorsinfluencing the likelihood of occurrence of S. parksii using boosted regression trees. Our model classified 97 % of the cellscorrectly with regard to species presence and absence, and indicated that probability of existence was correlated withclimatic conditions and landscape features. The most influential variables were mean annual precipitation, mean ele-vation, mean annual minimum temperature and mean annual maximum temperature. The most likely suitable range forS. parksii was the eastern portions of Leon and Madison Counties, the southern portion of Brazos County, a portion ofnorthern Grimes County and along the borders between Burleson and Washington Counties. Our model can assist inthe development of an integrated conservation strategy through: (i) focussing future survey and research efforts onareas with a high likelihood of occurrence, (ii) aiding in selection of areas for conservation and restoration and (iii) framingfuture research questions including those necessary for predicting responses to climate change. Our model could alsoincorporate new information on S. parksii as it becomes available to improve prediction accuracy, and our methodologycould be adapted to develop distribution maps for other rare species of conservation concern.

Keywords: Boosted regression trees; conservation; endangered species; Navasota ladies’ tresses; reintroduction;species distribution models.

* Corresponding author’s e-mail address: [email protected]† These authors contributed equally to this work.

Published by Oxford University Press on behalf of the Annals of Botany Company.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

AoB PLANTS www.aobplants.oxfordjournals.org & The Authors 2015 1

http://creativecommons.org/licenses/by/4.0/



IntroductionConservation biologists and natural resource managersare growing increasingly concerned about the mannerin which climate change and accelerating habitat frag-mentation may negatively affect the long-term viabilityof threatened and endangered plant species (Wilcoveet al. 1998; Pitman and Jorgensen 2002; Brigham andSwartz 2003). To successfully prevent their extirpation,conservation efforts will require detailed studies of spe-cies population biology and life-history dynamics, morethorough assessments of the factors contributing torarity, sophisticated land management and restorationstrategies and the development of more robust predictivemodels that better identify both high-priority conserva-tion locations as well as areas potentially suitable forplant reintroductions (e.g. Falk and Holsinger 1991;Schemske et al. 1994; Maschinski and Haskins 2012).

Orchidaceae is the largest and most diverse family offlowering plants, but it is currently facing unprecedentedrisks of extinction (Cribb et al. 2003; Swarts and Dixon2009). Orchidaceae consists of over 1000 genera andmost orchid genera contain one or more threatened orendangered species (Cribb et al. 2003). The majority ofthreatened orchid species are terrestrial orchids, despitethe small portion of the family represented by this lifeform (IUCN 2001). In addition, many terrestrial orchidsare rare, with specialized habitat requirements, makingthem particularly susceptible to habitat fragmentationand modification (Wu and Smeins 2000; Pillon andChase 2007). Their vulnerability is exacerbated by patchydistributions, specialized mutualisms and generally lim-ited dispersal (Schemske et al. 1994; Coates et al. 2006;Pillon and Chase 2007). Given the high extinction riskto terrestrial orchids, they have been a major conserva-tion concern for many environmental groups. They haveoften been used as flagship species in conservationinitiatives because of their uniqueness and rarity andadditionally are often touted as important early warningbioindicators for ecosystem health given their sensitivityto environmental degradation (Cribb et al. 2003; Swartsand Dixon 2009).

Despite substantial conservation emphasis on rareorchids, populations continue to decline (Swarts andDixon 2009). This is in no small part due to difficulties indesigning integrated conservation plans for orchid pro-tection (Whigham and Willems 2003; Swarts et al.2007). Often, small patchily distributed orchid popula-tions are difficult to detect without thorough surveys ofextensive areas, which are often not logistically feasible(Wu and Smeins 2000; Gogol-Prokurat 2011). In addition,many populations are spread across a network of privateor otherwise inaccessible land. Incomplete censuses of

populations limit the ability for conservation planners todetermine appropriate areas for protected habitat andassisted migrations (Cuperus et al. 1999; Wan et al.2014). Effective conservation planning requires the iden-tification of areas of suitable habitat in order to facilitateprioritization and appropriately identify land for the cre-ation of preserves or easements and mitigation forhabitat modification or loss (Rodrıguez et al. 2007; Gogol-Prokurat 2011; Dudley and Bean 2012).

Species distribution models are invaluable tools for focus-sing conservation efforts of species with incomplete distri-bution records (Fleishman et al. 2002; Buse et al. 2007;Fandohan et al. 2011; Gogol-Prokurat 2011). Species distri-bution models comprise a suite of quantitative tools thatstatistically relate species presence and absence data toenvironmental predictor variables (Guisan and Thuiller2005; Gogol-Prokurat 2011). They elucidate habitat require-ments, aiding in the development of distribution predictionsessential to meeting endangered species conservationobjectives with limited site occupancy data and resourcesfor additional data collection (Guisan and Thuiller 2005;Hirzel et al. 2006). They can be used to identify habitat suit-able for conservation by providing maps of probabilities thatthe species would occur in a given area (Ibisch et al. 2002;Jimenez-Valverde and Lobo 2007), determine the effects ofland-use change on endangered species habitat (Rodrıguezet al. 2007) and explore the effect of global change onendangered species distributions (Jimenez-Valverde andLobo 2007; Thuiller et al. 2008).

Studies involving species distribution modelling haveincreased in recent years and several methods are currentlyapplied to address ecological issues (Elith and Leathwick2009). These include statistical models such as generalizedlinear models (Wang and Grant 2012) and generalizedadditive models (Leathwick et al. 2006), machine-learningmodels such as CLIMEX (Pattison and Mack 2008), GARP(Stockman et al. 2006) and Maxent (Wilting et al. 2010),as well as methods drawing on insights and techniquesfrom statistical and machine learning approaches suchas random forests (Prasad et al. 2006) and boosted regres-sion trees (Chiou et al. 2013). Boosted regression trees area relatively new method compared to others. Boostedregression trees have their origins within machine learning,but subsequent developments in the statistical communityhave led to a reinterpretation of boosted regression trees asan advanced form of regression (Elith et al. 2008). However,boosted regression trees differ fundamentally from sta-tistical methods and machine-learning approaches suchas Maxent that produce a single ‘best’ model in thatboosted regression trees combine a large number of rela-tively simple tree models adaptively to optimize predictivepower (Leathwick et al. 2006). Each of the individualmodels consists of a simple classification or regression

2 AoB PLANTS www.aobplants.oxfordjournals.org & The Authors 2015

Wang et al. — Species distribution modelling for conservation of an endangered orchid

tree (a rule-based classifier) that partitions observationsinto groups having similar values for the response variablebased on a series of binary rules constructed from the in-dependent variables (Hastie et al. 2001). Boosted regres-sion trees have been used to predict the distribution of athreatened species, rabbitsfoot (Quadrula cylindrical), viathe inclusion of independent variables measured at mark-edly different spatial scales (Hopkins 2009).

In this study, we analysed the relationships betweenthe occurrence of Spiranthes parksii, an endangeredterrestrial orchid, and several climatic and landscape vari-ables deemed important to S. parksii distribution. Spir-anthes parksii is a state and federally listed endangeredorchid, endemic to central Texas, USA. Its distributionis limited to 13 counties and it appears to occupy arestricted habitat within those counties, often observedon the edges of upland drainages in small open grass/shrub patches within post-oak savannah/woodland com-munities (Wonkka et al. 2012). Spiranthes parksii popula-tions are threatened by land conversion to agricultureand lignite mines, urban development and woody en-croachment by trees and understory shrubs into post-oaksavannahs (Wonkka et al. 2012). Like many terrestrialorchids, S. parksii is a mycoheterotroph, requiring amycorrhizal symbiont for germination and seedlingdevelopment, remaining closely associated with thefungi throughout its life cycle (Ariza 2013). This specia-lized mutualism, along with a limited dispersal shadow,and specialized habitat requirements leads to a patchydistribution, making S. parksii detection difficult. In add-ition, the counties in which the populations are locatedconsist largely of privately owned land, inaccessible tosurveyors. High rates of development in this region neces-sitate effective prioritization of lands for mitigationefforts. Given the conservation concerns and limitedavailability of survey data for this species, we developeda species distribution model to aid in effective conserva-tion of S. parksii. In particular, we used boosted regressiontrees to (i) identify potential factors influencing S. parksiidistribution, (ii) quantify the relative importance of eachfactor and (iii) predict suitable S. parksii habitat. Themodel developed herein will provide an adaptive quanti-tative tool which can be used to facilitate future S. parksiisurveying, research and conservation efforts and, withslight modification, should be applicable to other endan-gered species with similarly limited ranges.

Methods

Study area and data sources

The study area covers several counties (Bastrop, Fayette,Milam, Freestone, Leon, Madison, Grimes, Robertson andBrazos) in central Texas, USA (Fig. 1). The area is largely

post-oak savannah intermixed with open grassland, crop-land and urban and suburban development. The climateis humid subtropical with an average minimum tempera-ture of 14 8C, an average maximum temperature of 26 8Cand average annual precipitation of 105 cm bimodallydistributed with peaks in the fall and the spring.

We obtained geo-referenced data on (i) presence andabsence of S. parksii regularly sampled between 2004and 2012 from the US Fish and Wildlife Service, TexasParks and Wildlife Department, Texas Department ofTransportation and the Texas A&M University team work-ing under Drs Smeins and Rogers, (ii) average climaticconditions in our study area from the PRISM Climategroup (2013) and (iii) landscape features including topo-graphic characteristics derived from digital elevationmodels (Gesch 2007), land cover (Fry et al. 2011), soilcharacteristics (Soil Data Mart 2013) and geology (Stoeseret al. 2013). Spiranthes parksii surveys were conduc-ted yearly during peak flowering for the duration of thestudy with systematic sampling across areas of knownS. parksii occupancy and additional areas deemed likelyto contain S. parksii due to similarity in habitat character-istics to known areas of occupancy. Spiranthes parksiilocations were marked with GPS coordinates for futureobservation.

Our geo-referenced data for S. parksii were primarilycomprised of occurrences. Pseudo-absences, or randompoints in the study area where the focal species has notbeen documented, are typically used as a surrogate forabsence records in studies with such data limitations.However, there may be limited confidence in absenceat those points, depending on the sampling strategy(Phillips et al. 2009). As a more definitive sample of ab-sence points, we used locality records for an ecologicallysimilar congener, S. cernua. Given the similar ecology andphenology of the two species (Ariza 2013), and the con-servation status of S. parksii, S. parksii would have beenrecorded if found during surveys for S. cernua. Thus, weconfidently use records for S. cernua that lack concurrentrecords for S. parksii.

Data analysis

We selected 58 variables that have been suggested in theliterature as potential predictors of the presence ofS. parksii in central Texas (Wonkka et al. 2012; Krupnicket al. 2013), including various climatic conditions andlandscape features [see Supporting Information].We analysed relationships between the occurrence ofS. parksii and the potential explanatory variables byaggregating the explanatory variable data associatedwith S. parksii presence (106 cells) and absence (99cells) into polygons representing a resolution of 800 ×800 m cells, aligned with the climate data that we used,



http://aobpla.oxfordjournals.org/lookup/suppl/doi:10.1093/aobpla/plv039/-/DC1

in central Texas. We then merged these data into a gridof 37 427 800 × 800 m cells using ArcGIS 9.0 (ESRI2009). The climate data included annual average max-imum and minimum temperatures, and precipitation for1981–2010 (PRISM Climate Group 2013). We derivedtopographic characteristics for the 800 m grid cellsof analysis from the 30 m National Elevation Dataset(Gesch 2007) using SAGA GIS version 2.1.0 (www.sagagis.org). We also calculated the average soil water-holding capacity, percentage of silt, sand and clay ineach soil type based on STATSGO soil data (Soil DataMart 2013) using R version 3.0.2 (R Core Team 2013). Weused spatial overlay tools in SAGA GIS version 2.1.0 andManifold GIS version 8.0.28 to aggregate the variousdata layers [see Supporting Information] into a singledataset for analyses.

We conducted our analysis using boosted regressiontrees which combine decision trees and a boosting algo-rithm with a form of logistic regression (Friedman 2002;De’ath 2007; Elith et al. 2008). For boosted regressiontrees, the probability (P) of S. parksii occurrence (y ¼ 1)at a location with the potential explanatory variables(X ) is given by P(y ¼ 1|X ) and is modelled via thelogit: logit P(y ¼ 1|X ) ¼ f (X ). We fitted our model in

R (R Development Core Team 2006 version 2.14.1) usingthe gbm package version 1.5-7 (Ridgeway 2006) andcode provided by Elith et al. (2008). The optimal modelwas determined following the recommendations of Elithet al. (2008) by altering the learning rate and tree com-plexity (the number of split nodes in a tree) until the pre-dictive deviance was minimized without over-fitting, andby limiting our choice of the final model to those that con-tained at least 1000 trees (where each successive tree isbuilt for the prediction residuals of the preceding tree).Once the optimal combination of learning rate and treecomplexity was found, model performance was evalu-ated using a 10-fold cross-validation procedure withresubstitution. For each cross-validation trial, 80 % ofthe dataset was randomly selected for model fittingand the excluded 20 % was used for testing. We calcu-lated the response variance explained, the area underthe receiver operator characteristic curve (AUC), the over-all accuracy, the omission error rate and the commissionerror rate based on the aggregated CV results. We evalu-ated the reliability and validity of our models as fair(0.50 , AUC ≤ 0.75), good (0.75 , AUC ≤ 0.92), verygood (0.92 , AUC ≤ 0.97), or excellent (0.97 , AUC ≤1.00) based on the value of AUC (Hosmer and Lemeshow

Figure 1. The study area and the current distribution (filled circles) of the endangered orchid S. parksii in central Texas, USA.



http://aobpla.oxfordjournals.org/lookup/suppl/doi:10.1093/aobpla/plv039/-/DC1

2000). We then used the gbm library to derive the relativeinfluence of each potential explanatory variable in themodel and constructed partial dependence plots for themost influential variables (Elith et al. 2008). Finally, weused this optimal model to calculate probability ofS. parksii presence in each cell in central Texas and super-imposed these probabilities of occupancy on a map of thestudy area using ArcMap 9.0 (ESRI 2009).

ResultsAnalyses of 500 combinations of tree complexity (rangingfrom 3 to 7) and learning rate (ranging from 0.001 to 0.01)produced models with between 450 and 3900 treeswhose values of predictive deviance ranged from 0.582to 0.624. The optimal model had a tree complexity of 5,a learning rate of 0.003 and a total of 1200 trees. Modelpredictive deviance was 0.582+0.0079 with 95.6 % ofthe total response variance explained. The AUC scorewas 0.940+0.016 (‘very good’ ability to discriminatebetween species presence and absence) and the overallaccuracy was 91.7 %. The commission (false positive)error rate was 6.8 % and the omission (false negative)error rate was 9.8 %. Recursive feature elimination testsshowed that 45 variables could be removed from themodel before the resulting predictive deviance exceededthe initial predictive deviance of the model with allvariables.

Thirteen variables were included in the final model(Table 1), with variables associated with climatic

conditions and landscape features accounting for �53.5and 46.5 %, respectively, of the contribution in the overallmodel (Fig. 2). Examination of the relative contributionof the predictor variables indicated that the top fouraccounted for �70.95 % of the contribution in the overallmodel. Of the four most influential model variables, threewere climatic conditions and one was a landscape fea-ture. Mean annual precipitation, mean annual minimumtemperature and mean annual maximum temperaturewere the first, third and fourth most influential variables,contributing 26.93, 17.26 and 9.31 %, respectively. Meanelevation was the second most important variable con-tributing 17.45 %.

Partial dependence plots indicated that S. parksii occur-rences were associated with climatic conditions charac-terized by mean annual precipitation between 1050 and1120 mm (Fig. 3A), mean annual minimum temperaturebetween 13.75 and 14.38 8C (Fig. 3C) and mean annualmaximum temperature between 26.00 and 26.25 8C(Fig. 3D). Occurrences also were associated with land-scape features characterized by (i) an altitude between50 and 80 m (Fig. 3B), (ii) a slope ratio between 0.05and 0.09 % (Fig. 3H), (iii) areas with ,20 % pasture(Fig. 3E), 20–73 % of evergreen forest (Fig. 3I), 50–73 %of deciduous forest (Fig. 3J), or ,5 % of developed openspace (Fig. 3K), (iv) soil with ,20 % clay (Fig. 3F) or 55–94 % sand (Fig. 3L) and (v) geological formations inwhich .75 % belonged to the Manning formation(Fig. 3G) or in which .40 % belonged to the Wellbornformation (Fig. 3M).

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 1. Abbreviations, descriptions and descriptive statistics for the climatic conditions and landscape features included in the final model.

Variable Description Mean Minimum Maximum

Climatic conditions

pptCrop.13 Mean annual precipitation (mm × 100) 104 742 88 153 115 293

TMinCrop.13 Mean annual minimum temperature (C × 100) 1372 1255 1438

Mean annual maximum temperature (C × 100) 2598 2499 2671

Landscape features

DEM.Mean Mean elevation (m) 101.49 48.40 194.61

Pasture.Hay Proportion of pasture (%) 0.36 0 0.97

STATSGO_AvgClay Percentage of clay based on average of soil types (%) 28.27 5.04 51.91

TXEOm Percentage of Manning formation on average of geological formation (%) 0.29 0 1

Slope.Mean Mean slope (degree × 100) 0.03 0 0.09

Evergreen.Forest Proportion of evergreen forest (%) 0.05 0 0.73

Deciduous.Forest Proportion of deciduous forest (%) 0.14 0 0.73

Developed.Open.Space Percentage of developed open space (%) 0.04 0 0.61

STATSGO_AvgSand Percentage of sand based on average of soil types (%) 42.57 15.53 94.30

TXEOwb Percentage of Wellborn formation on average of geological formation (%) 0.06 0 1



Our analyses suggest that potential habitat for S. parksiiin central Texas, considering its association with the vari-ables mentioned in the previous paragraph, is most likelyto be (i) the eastern portions of Leon and Madison Coun-ties, (ii) the southern portion of Brazos County, (iii) a por-tion of northern Grimes County and (iv) along the bordersbetween Burleson and Washington Counties (Fig. 4).

Approximately 84, 5, 4, 3, 3 and 1 % of the cells fell withinthe P ≤ 0.5, 0.5 , P ≤ 0.6, 0.6 , P ≤ 0.7, 0.7 , P ≤ 0.8,0.8 , P ≤ 0.9 and 0.9 , P ≤ 1.0 estimated probability ofoccurrence (P) categories, respectively.

DiscussionPlant distributions are limited by the availability of suitablehabitats (Aitken et al. 2007). For rare plants, especiallythose with limited geographic ranges, narrow habitatspecificity can further limit distribution. While climate isan important determinant of plant distribution at land-scape levels (Pearson et al. 2004), soil properties and bioticinteractions determine habitat availability at local scales(Raven 2002). Even for edaphic endemics, combinationsof variables predict distributions more accurately thansimple models driven entirely by soil-related parameters(Arundel 2005). This is especially valid for predicting orchidspecies distributions which are highly dependent on in-teractions with pollinators and mycorrhizal symbionts(Rasmussen 2002; Rasmussen and Rasmussen 2009).

Our model establishes the importance of both climaticvariables and landscape features to the distribution ofS. parksii. Spiranthes parksii is associated with the higher

Figure 3. Partial dependence plots for the 13 most influential variables included in the final model. The y-axis represents the logit scale used forthe indicated variable, hash marks at the top of the plot indicate the locations of the sample sites along the range of the variables.

Figure 2. Relative contributions (%) of the 13 most influentialvariables included in the final model (see Table 1 for the descriptionof variables).



end of the range of average annual precipitation forthe area (1050–1120 mm). This agrees with findings byAriza (2013) that showed higher soil moisture as a majorexplanatory variable differentiating S. parksii occurrencewith the more abundant sympatric species S. cernua.Spiranthes parksii is also found in areas with high minimum(13.8–14.4 8C) and maximum (26–26.3 8C) mean annualtemperatures. This likely contributes to S. parksii distribu-tion through unique life-history characteristics includingsummer dormancy and potential early fall emergence ofrosettes (Wonkka et al. 2012). Summer dormant plantsexisting below the soil as rhizomes can withstand highpeak temperatures, but with above-ground photosynthetic

vegetation being present in the winter, the plants favourareas with higher winter temperatures to minimize poten-tial frost damage.

Given the importance of climatic variables to S. parksiidistribution, climate change could have an extensiveimpact on the availability of suitable S. parksii habitatin the future. Climate change has been shown to causedistribution shifts for many species of plants and canincrease the likelihood of local extinction as sessile plantspecies are unable to disperse or adapt to a rapidly chan-ging climate (Hoegh-Guldberg et al. 2008; Nicole et al.2011; Parmesan et al. 2013). This is further exacerbatedin highly fragmented areas, such as the range of S. parksii,

Figure 4. Estimated probabilities of occurrence of S. parksii in central Texas, USA.



where human alteration can act as a barrier to dispersalprocesses (Preston et al. 2008). Although there is consider-able uncertainty in predictions of future temperatures andprecipitation for a particular region, projections averagedacross ensemble models suggest increased summer andwinter temperatures and decreased average annual precipi-tation across the southern United States (Deser et al. 2012).In addition, precipitation is expected to become more vari-able, with more frequent drought events and more precipi-tation occurring in fewer rainfall events (Coumou andRahmstorf 2012; Deser et al. 2012; IntergovernmentalPanel on Climate Change 2014). While warmer tempera-tures could increase the habitat available to S. parksii,there is likely an upper bound on temperatures that theorchid can withstand. In addition, reduced precipitationand greater frequency of drought could cause many cur-rently suitable areas of habitat to become too dry to supportpopulations of the orchid given that our model showsS. parksii to occur in the wetter portions of its range.

Several landscape features also proved important toS. parksii distribution. Elevation and slope are also im-portant to determining S. parksii occurrence. Althoughslope and elevation are not mechanistic variables, theyoften can be proxies for environmental variables, suchas soil properties and plant-available water, which candrive plant distributions (Lassueur et al. 2006). Elevation,derived from digital elevation models, is the most import-ant landscape feature for predicting S. parksii distribution.They are found at the low end of the elevation range forthe area (50–80 m). This is reflective of the specific habi-tat preference for margins of drainages (Ariza 2013).Similarly, S. parksii occur in areas with maximum sloperatios for the area (0.05–0.09), which also reflects theiroccurrence between flatter open areas and margins ofdrainages.

Our model also showed soils and vegetation cover typeto be important for the distribution of S. parksii. Thisis consistent with past studies that found high S. parksiioccurrence in the Manning and Wellborn geologicalformations, suggesting that S. parksii might be an edaphicendemic. This is also consistent with the life cycle depend-ence on mycorrhizal fungi (Ariza 2013). Orchid distributionsare thought to be restricted largely by interactions withpollinators and their mycorrhizal symbionts (Watermanand Bidartondo 2008). For S. parksii, pollination is likelyless important (as evidenced by high levels of asexualreproduction) than fungal mutualism (Ariza 2013). Fungitend to be patchily distributed across a landscape (Battyet al. 2001), and their distributions are driven largely bylocal mechanisms, especially soil properties such as soilmoisture and soil organic matter (Brundrett and Abbott1994). Ariza (2013) found higher summer soil moisture,lower pH, percent sand and abundance of soil organic

matter to be the most important distinguishing character-istics between S. parksii and S. cernua occurrence. Ourmodel suggests that S. parksii soils usually are found inareas with ,20 % clay and 55–94 % sand. The importanceof soil organic matter to S. parksii likely is related to boththe ability of organic matter to increase the water-holdingcapacity of well-drained sandy soils, and also its import-ance as a substrate for fungi. Orchids also tend to exhibitnarrow specificity with fungi (Waterman and Bidartondo2008; Rasmussen and Rasmussen 2009). Therefore, it islikely that S. parksii distribution closely tracks particularfungal species. The distribution of those fungi is likely dri-ven by specific soil inputs as well as soil properties, whichcould explain the importance of vegetative cover (,20 %pasture, 20–73 % evergreen, 50–73 % deciduous) toS. parksii distribution. Fungi likely require leaf litter as asubstrate for decomposition. However, a thick layer ofleaf litter might inhibit germination of S. parksii seeds.This is supported by the findings of Ariza (2013) that uni-form leaf litter cover was an important determinant ofS. parksii occurrence. Soils have been found to be importantdeterminants of distribution for other orchids. Bowles et al.(2005) found soil type to be the most important variabledetermining Plantanthera leucophaea distribution andClark et al. (2004) determined climate, soils and vegetationtype to accurately predict distributions of Cryptostylishunteriana.

Areas with high estimated probabilities of S. parksiioccurrence are distributed patchily across the range.There are some larger connected areas of high probability,but many areas with high likelihood of occurrence arepunctuated with lower likelihood patches. Plants withlimited dispersal shadows are highly susceptible to localextinctions due to stochastic events in fragmentedhabitats (Bourg et al. 2005). The distribution map gener-ated with our model suggests fragmented habitat forS. parksii, which has limited dispersal due to tiny seedsand the necessity for mycorrhizal associations for germin-ation. Many seedlings are found in close proximity to adultplants (Ariza 2013). Patchy distributions often are asso-ciated with limited habitat availability (Swarts and Dixon2009). However, if there is little chance for long distancedispersal and patchily distributed areas of suitable habitat(Hurtt and Pacala 1995), movement into unoccupied areasof suitable habitat could be restricted, posing problems forspecies with high frequencies of local extinctions (Primackand Miao 1992). This is especially significant in areas thatare highly fragmented by development and agriculturesuch as the central Texas range of S. parksii. Additionally,distribution shifts necessary for species continuation inthe face of climate change require areas of suitable habitatbe attainable for future recruitment and persistence (Careyand Brown 1994). Existence of such sites becomes



increasingly less likely with high habitat specificity, limiteddispersal distances and fragmented habitat. Our modelpossesses potential utility for understanding S. parksiimeta-population dynamics more thoroughly to determinethe level of threat posed to species viability in the face ofincreased landscape fragmentation.

The model developed in this study has potential utilitybeyond the scope of this work. It can be adaptedto incorporate new information and data as they be-come available. Model accuracy increases with increasedamount and accuracy of presence and absence data andcan be updated to include new information to further re-fine distribution predictions (Elith and Leathwick 2009).Additionally, modelling multiple scales could increasethe accuracy of prediction. Ecological processes functionat different scales (Turner 1989; Levin 1992). Our modelexplores landscape level scales that drive distribution,but refining the model resolution could yield importantinformation regarding distribution at a finer scale (i.e.within a high probability patch). Fine scale mechanismsoften regulate the distribution of rare plants with specia-lized habitat requirements (Menges et al. 1999). Forinstance, Diez and Pulliam (2007) found that distancefrom parent was important to germination at the localscale, while soil characteristics were more predictive ofgermination at larger scales for Goodyera pubescencs.One opportunity for improvement of the model is toincorporate data related to disturbance and biotic inter-actions (e.g. distribution of pollinators or fungal associ-ates, fire or flooding effects on habitat quality) in orderto reflect the potential for non-equilibrium system func-tioning (Schroder and Seppelt 2006). This could proveespecially important for species such as rare orchidsthat have specific biotic interactions (Ettema and Wardle2002) and tend to respond to specific disturbance regimes(Clark et al. 2004). Models incorporating landscape changesand mechanistic drivers can better capture fluctuations inhabitat suitability over time (Kearney et al. 2008). Thiscould increase the accuracy of distribution predictions insystems where habitat quality fluctuates in response tonon-equilibrium processes (Elith and Leathwick 2009).

ConclusionsA suite of climatic variables and landscape features can beused to predict the distribution of the endangered terres-trial orchid, S. parksii which is endemic to central Texas.Many of these variables are related to soil resourceswhich potentially influence the distribution of the mycor-rhizal fungi the orchid depends on for germination and life-time nutrient acquisition (Rasmussen and Rasmussen2007; Ariza 2013). The species’ potential habitat is patchilydistributed as a result of this dependence on soil resources

and specific habitat requirements (Batty et al. 2001).Narrow habitat specificity combined with potential disper-sal limitations necessitates an integrated conservationapproach that includes research to determine basic eco-logical and biological processes important to S. parksiipopulation viability, habitat management and conserva-tion and an understanding of the effects of fragmentationand habitat degradation on dispersal of S. parksii into suit-able habitat (Swarts and Dixon 2009). Species distributionmodels can assist in the development of an integrated con-servation strategy (Kiesecker et al. 2010). They can help tofill knowledge gaps resulting from limited resources forresearch. Similarly, they can help focus future survey andresearch efforts on areas with a high likelihood of occur-rence (Parris 2002; Guisan et al. 2006). Species distributionmodels also can be used to select areas for conservationoffsets or easements (Gibbons and Lindenmayer 2007;Kumar and Stohlgren 2009), explore alternate manage-ment scenarios (Guisan et al. 2013), frame researchquestions (Aitken et al. 2007), explore issues related tometa-population dynamics (Bourg et al. 2005) and predictpotential responses to climate change (Carey and Brown1994). The species distribution model developed throughthis research is adaptive. It can incorporate new informa-tion as it becomes available to improve accuracy and reso-lution of our analyses. Our methodology could also beemployed to develop distribution maps for other rare spe-cies of conservation concern.

Sources of FundingThis research was funded by the City of Bryan/CollegeStation-Brazos Valley Solid Waste Management Agency,Texas Department of Transportation and a Ladybird John-son Wildflower Center (Austin, TX)—Endangered SpeciesConservation Grant Program Award #12419. C.L.W. wassupported by a USDA-CSREES National Needs Fellowship(2009-38420-05631) and M.L.T. was supported by anNSF-IGERT Traineeship through the Applied BiodiversityScience Program at Texas A&M University (NSF DGE0654377). C.L.W. and M.L.T. were also supported by theTom Slick Doctoral Fellowship Program.

Contributions by the AuthorsAll authors shared in collecting data, constructing themodel and writing.

Conflict of Interest StatementNone declared.



AcknowledgementsWe thank M. Ariza, D. Nally, J.R. Hammons, R. Bruton, S.J.Haller, A. Richards, A. Delgado and many Texas A&M Uni-versity undergraduate research students who providedassistance with field work. We also acknowledge S. Bestand BVSWMA staff for logistical assistance with field studies.

Supporting InformationThe following additional information is available in theonline version of this article –

Appendix. Abbreviations, descriptions and descriptivestatistics for the 94 climatic conditions and landscapefeatures identified as potential factors influencing thelikelihood of existence of S. parksii in the central Texas,USA.

Literature CitedAitken M, Roberts DW, Shultz LM. 2007. Modeling distributions of rare

plants in the Great Basin, western North America. Western NorthAmerican Naturalist 67:26–38.

Ariza MC. 2013. Mycorrhizal associations, life history, and habitatcharacteristics of the endangered terrestrial orchid Spiranthesparksii corell and sympatric congener Spiranthes cernua: implica-tions for conservation. PhD dissertation, Texas A&M University,College Station, TX.

Arundel ST. 2005. Using spatial models to establish climatic limitersof plant species’ distributions. Ecological Modelling 182:159–181.

Batty AL, Dixon KW, Brundrett M, Sivasithamparam K. 2001. Con-straints to symbiotic germination of terrestrial orchid seed in amediterranean bushland. New Phytologist 152:511–520.

Bourg NA, McShea WJ, Gill DE. 2005. Putting a CART before thesearch: successful habitat prediction for a rare forest herb.Ecology 86:2793–2804.

Bowles M, Zettler L, Bell T, Kelsey P. 2005. Relationships between soilcharacteristics, distribution and restoration potential of thefederal threatened eastern prairie fringed orchid, Platantheraleucophaea (Nutt.) Lindl. The American Midland Naturalist 154:273–285.

Brigham CA, Swartz MA (eds) 2003. Population viability in plants:conservation, management, and modeling of rare plants. Berlin:Springer.

Brundrett MC, Abbott LK. 1994. Mycorrhizal fungus propagules in theJarrah forest. I. Seasonal study of inoculum levels. New Phytolo-gist 127:539–546.

Buse J, Schroder B, Assmann T. 2007. Modelling habitat and spatial dis-tribution of an endangered longhorn beetle—A case study forsaproxylic insect conservation. Biological Conservation 137:372–381.

Carey PD, Brown NJ. 1994. The use of GIS to identify sites that willbecome suitable for a rare orchid, Himantoglossum hircinum L.,in a future changed climate. Biodiversity Letters 2:117–123.

Chiou C-R, Wang H-H, Chen Y-J, Grant WE, Lu M-L. 2013. Modelingpotential range expansion of the invasive shrub Leucaena leuco-cephala in the Hengchun peninsula, Taiwan. Invasive PlantScience and Management 6:492–501.

Clark S, deLacey C, Chamberlain S. 2004. Using environmental vari-ables and multivariate analysis to delineate preferred habitatfor Cryptostylis hunteriana, the leafless tongue orchid, in theShoalhaven local government area, NSW. Cunninghamia 8:467–476.

Coates F, Lunt ID, Tremblay RL. 2006. Effects of disturbance on popu-lation dynamics of the threatened orchid Prasophyllum correc-tum D.L. Jones and implications for grassland management insouth-eastern Australia. Biological Conservation 129:59–69.

Coumou D, Rahmstorf S. 2012. A decade of weather extremes.Nature Climate Change 2:491–496.

Cribb PJ, Kell SP, Dixon KW, Barrett RL. 2003. Orchid conservation: aglobal perspective. In: Dixon KW, Kell SP, Barrett RL, Cribb PJ, eds.Orchid conservation. Kota Kinabalu, Sabah: Natural HistoryPublications, 1–24.

Cuperus R, Canters KJ, Udo de Haes HA, Friedman DS. 1999. Guide-lines for ecological compensation associated with highways.Biological Conservation 90:41–51.

De’ath G. 2007. Boosted trees for ecological modeling and prediction.Ecology 88:243–251.

Deser C, Knutti R, Solomon S, Phillips AS. 2012. Communication of therole of natural variability in future North American climate.Nature Climate Change 2:775–779.

Diez JM, Pulliam HR. 2007. Hierarchical analysis of species distribu-tions and abundance across environmental gradients. Ecology88:3144–3152.

Dudley TL, Bean DW. 2012. Tamarisk biocontrol, endangered speciesrisk and resolution of conflict through riparian restoration.BioControl 57:331–347.

Elith J, Leathwick JR. 2009. Species distribution models: ecologicalexplanation and prediction across space and time. AnnualReview of Ecology, Evolution, and Systematics 40:677–697.

Elith J, Leathwick JR, Hastie T. 2008. A working guide to boostedregression trees. Journal of Animal Ecology 77:802–813.

ESRI. 2009. ArcGIS. Redlands, CA: Environmental Systems ResearchInstitute.

Ettema CH, Wardle DA. 2002. Spatial soil ecology. Trends in Ecologyand Evolution 17:177–183.

Falk DA, Holsinger KE (eds) 1991. Genetics and conservation of rareplants. New York: Oxford University Press.

Fandohan B, Assogbadjo AE, Glele Kakaı RL, Sinsin B. 2011.Effectiveness of a protected areas network in the conservationof Tamarindus indica (Leguminosea-Caesalpinioideae) in Benin.African Journal of Ecology 49:40–50.

Fleishman E, Ray C, Sjogren-Gulve P, Boggs CL, Murphy DD. 2002.Assessing the roles of patch quality, area, and isolation inpredicting metapopulation dynamics. Conservation Biology 16:706–716.

Friedman JH. 2002. Stochastic gradient boosting. ComputationalStatistics and Data Analysis 38:367–378.

Fry JG, Xian SJ, Dewitz J, Homer C, Yang L, Barnes C, Herold N,Wickham J. 2011. Completion of the 2006 national land coverdatabase for the conterminous United States. PhotogrammetricEngineering and Remote Sensing 77:858–864.

Gesch DB. 2007. The national elevation dataset. In: Maune D, ed.Digital elevation model technologies and applications: the DEMuser’s manual, 2nd edn. Bethesda, MD: American Society forPhotogrammetry and Remote Sensing, 99–118.



Gibbons P, Lindenmayer DB. 2007. Offsets for land clearing: no netloss or the tail wagging the dog? Ecological Management andRestoration 8:26–31.

Gogol-Prokurat M. 2011. Predicting habitat suitability for rare plantsat local spatial scales using a species distribution model.Ecological Applications 21:33–47.

Guisan A, Thuiller W. 2005. Predicting species distribution: offering morethan simple habitat models. Ecology Letters 8:993–1009.

Guisan A, Broennimann O, Engler R, Vust M, Yoccoz NG, Lehmann A,Zimmermann NE. 2006. Using niche-based models to improvethe sampling of rare species. Conservation Biology 20:501–511.

Guisan A, Tingley R, Baumgartner JB, Naujokaitis-Lewis I, Sutcliffe PR,Tulloch AIT, Regan TJ, Brotons L, McDonald-Madden E, Mantyka-Pringle C, Martin TG, Rhodes JR, Maggini R, Setterfield SA, Elith J,Schwartz MW, Wintle BA, Broennimann O, Austin M, Ferrier S,Kearney MR, Possingham HP, Buckley YM. 2013. Predicting spe-cies distributions for conservation decisions. Ecology Letters 16:1424–1435.

Hastie T, Tibshirani R, Friedman JH. 2001. The elements of statisticallearning: data mining, inference, and prediction. New York, NY:Springer.

Hirzel AH, Le Lay G, Helfer V, Randin C, Guisan A. 2006. Evaluating theability of habitat suitability models to predict species presences.Ecological Modelling 199:142–152.

Hoegh-Guldberg O, Hughes L, McIntyre S, Lindenmayer DB,Parmesan C, Possingham HP, Thomas CD. 2008. Assisted colon-ization and rapid climate change. Science 321:345–346.

Hopkins RL. 2009. Use of landscape pattern metrics and multiscaledata in aquatic species distribution models: a case study of afreshwater mussel. Landscape Ecology 24:943–955.

Hosmer DW, Lemeshow S. 2000. Applied logistic regression. New York,NY: John Wiley and Sons, Inc.

Hurtt GC, Pacala SW. 1995. The consequences of recruitment limita-tion: reconciling chance, history and competitive differencesbetween plants. Journal of Theoretical Biology 176:1–12.

Ibisch PL, Nowicki C, Muller R, Araujo N. 2002. Methods for the assess-ment of habitat and species conservation status in data-poorcountries–case study of the Pleurothallidinae (Orchidaceae) ofthe Andean rain forests of Bolivia. In: Bussman RW, Lange S,eds. Conservation of biodiversity in the Andes and the Amazon.Proceedings of the Andes and the Amazon Basin Conference,Cusco, Peru. INKA e.V., Munich, Germany, 225–246.

Intergovernmental Panel on Climate Change. 2014. Climate change2013: the physical science basis: working group I contribution tothe fifth assessment report of the intergovernmental panel onclimate change. Cambridge: Cambridge University Press.

IUCN. 2001. IUCN Red List categories and criteria: version 3.1.Prepared by the IUCN Species Survival Commission.

Jimenez-Valverde A, Lobo JM. 2007. Threshold criteria for conversionof probability of species presence to either—or presence—absence. Acta Oecologica 31:361–369.

Kearney M, Phillips BL, Tracy CR, Christian KA, Betts G, Porter WP.2008. Modelling species distributions without using speciesdistributions: the cane toad in Australia under current and futureclimates. Ecography 31:423–434.

Kiesecker JM, Copeland H, Pocewicz A, McKenney B. 2010. Developmentby design: blending landscape-level planning with the mitigationhierarchy. Frontiers in Ecology and the Environment 8:261–266.

Krupnick GA, McCormick MK, Mirenda T, Whigham DF. 2013. The sta-tus and future of orchid conservation in north America. Annals ofthe Missouri Botanical Garden 99:180–198.

Kumar S, Stohlgren TJ. 2009. Maxent modeling for predicting suitablehabitat for threatened and endangered tree Canacomyricamonticola in New Caledonia. Journal of Ecology and the NaturalEnvironment 1:94–98.

Lassueur T, Joost SP, Randin CF. 2006. Very high resolution digital ele-vation models: do they improve models of plant species distribu-tion? Ecological Modelling 198:139–153.

Leathwick JR, Elith J, Hastie T. 2006. Comparative performance ofgeneralized additive models and multivariate adaptive regres-sion splines for statistical modelling of species distributions.Ecological Modelling 199:188–196.

Levin SA. 1992. The problem of pattern and scale in ecology: theRobert H. MacArthur Award Lecture. Ecology 73:1943–1967.

Maschinski J, Haskins KE (eds) 2012. Plant reintroduction in a chan-ging climate: promises and perils. Washington, DC: Island Press.

Menges ES, McIntyre PJ, Finer MS, Goss E, Yahr R. 1999. Microhabitatof the narrow Florida scrub endemic Dicerandra christmanii, withcomparisons to its congener D. frutescens. Journal of the TorreyBotanical Society 126:24–31.

Nicole F, Dahlgren JP, Vivat A, Till-Bottraud I, Ehrlen J. 2011.Interdependent effects of habitat quality and climate on popula-tion growth of an endangered plant. Journal of Ecology 99:1211–1218.

Parmesan C, Burrows MT, Duarte CM, Poloczanska ES, Richardson AJ,Schoeman DS, Singer MC. 2013. Beyond climate change attribu-tion in conservation and ecological research. Ecology Letters 16:58–71.

Parris KM. 2002. More bang for your buck: the effect of caller position,habitat and chorus noise on the efficiency of calling in the springpeeper. Ecological Modelling 156:213–224.

Pattison RR, Mack RN. 2008. Potential distribution of the invasive treeTriadica sebifera (Euphorbiaceae) in the United States: evaluatingclimex predictions with field trials. Global Change Biology 14:813–826.

Pearson RG, Dawson TP, Liu C. 2004. Modelling species distributionsin Britain: a hierarchical integration of climate and land-coverdata. Ecography 27:285–298.

Phillips SJ, Dudık M, Elith J, Graham CH, Lehmann A, Leathwick J,Ferrier S. 2009. Sample selection bias and presence-only distribu-tion models: implications for background and pseudo-absencedata. Ecological Applications 19:181–197.

Pillon Y, Chase MW. 2007. Taxonomic exaggeration and its effects onorchid conservation. Conservation Biology 21:263–265.

Pitman NCA, Jorgensen PM. 2002. Estimating the size of the world’sthreatened flora. Science 298:989.

Prasad AM, Iverson LR, Liaw A. 2006. Newer classification and regres-sion tree techniques: bagging and random forests for ecologicalprediction. Ecosystems 9:181–199.

Preston KL, Rotenberry JT, Redak RA, Allen MF. 2008. Habitat shifts ofendangered species under altered climate conditions: import-ance of biotic interactions. Global Change Biology 14:2501–2515.

Primack RB, Miao SL. 1992. Dispersal can limit local plant distribution.Conservation Biology 6:513–519.

PRISM Climate Group. 2013. Corvallis, OR: Oregon State University.http://prism.oregonstate.edu.



Rasmussen HN. 2002. Recent developments in the study of orchidmycorrhiza. Plant and Soil 244:149–163.

Rasmussen HN, Rasmussen FN. 2007. Trophic relationships in orchidmycorrhiza-diversity and implications for conservation. Lankes-teriana 7:334–341.

Rasmussen HN, Rasmussen FN. 2009. Orchid mycorrhiza: implica-tions of a mycophagous life style. Oikos 118:334–345.

Raven PH. 2002. Predicting species occurrences: issues of accuracyand scale. Washington, DC.

R Core Team. 2013. R: a language and environment for statisticalcomputing. Vienna, Austria: R Foundation for Statistical Comput-ing. http://www.R-project.org/.

Ridgeway G. 2006. Generalized boosted regression models.Documentation on the R package ‘gbm’, version 1-5-7. http://www.i-pensieri.com/gregr/gbm.shtml.

Rodrıguez JP, Brotons L, Bustamante J, Seoane J. 2007. The applica-tion of predictive modelling of species distribution to biodiversityconservation. Diversity and Distributions 13:243–251.

Schemske DW, Husband BC, Ruckelshaus MH, Goodwillie C, Parker IM,Bishop JG. 1994. Evaluating approaches to the conservation ofrare and endangered plants. Ecology 75:584–606.

Schroder B, Seppelt R. 2006. Analysis of pattern–process interactionsbased on landscape models—Overview, general concepts, andmethodological issues. Ecological Modelling 199:505–516.

Soil Data Mart. 2013. U.S. General soil map (STATSGO2). USDA/NRCS.http://soildatamart.nrcs.usda.gov.

Stockman AK, Beamer DA, Bond JE. 2006. An evaluation of a GARPmodel as an approach to predicting the spatial distribution of non-vagile invertebrate species. Diversity and Distributions 12:81–89.

Stoeser DB, Shock N, Green GN, Dumonceaux GM, Heran WD. 2013.A digital geologic map database for the state of Texas: U.S.Geological Survey data series. Denver, CO: U.S. Geological Survey.

Swarts ND, Dixon KW. 2009. Terrestrial orchid conservation in the ageof extinction. Annals of Botany 104:543–556.

Swarts ND, Batty AL, Hopper S, Dixon K. 2007. Does integrated con-servation of terrestrial orchids work? Lankesteriana 7:219–222.

Thuiller W, Albert C, Araujo MB, Berry PM, Cabeza M, Guisan A,Hickler T, Midgley GF, Paterson J, Schurr FM, Sykes MT,Zimmermann NE. 2008. Predicting global change impacts onplant species’ distributions: future challenges. Perspectives inPlant Ecology, Evolution and Systematics 9:137–152.

Turner MG. 1989. Landscape ecology: the effect of pattern on pro-cess. Annual Review of Ecology and Systematics 20:171–197.

Wan J, Wang C, Han S, Yu J. 2014. Planning the priority protectedareas of endangered orchid species in northeastern China. Bio-diversity and Conservation 23:1395–1409.

Wang H-H, Grant WE. 2012. Determinants of Chinese and Europeanprivet (Ligustrum sinense and Ligustrum vulgare) invasion andlikelihood of further invasion in southern U.S. forestlands. Inva-sive Plant Science and Management 5:454–463.

Waterman RJ, Bidartondo MI. 2008. Deception above, deceptionbelow: linking pollination and mycorrhizal biology of orchids.Journal of Experimental Botany 59:1085–1096.

Whigham DF, Willems JH. 2003. Demographic studies and life-history strategies of temperate terrestrial orchids as a basis forconservation. In: Dixon KW, Kell SP, Barrett RL, Cribb PJ, eds. Or-chid conservation. Kota Kinabalu, Borneo: Natural History Publi-cations, 137–158.

Wilcove DS, Rothstein D, Dubow J, Phillips A, Losos E. 1998. Quantify-ing threats to imperiled species in the United States. BioScience48:607–615.

Wilting A, Cord A, Hearn AJ, Hesse D, Mohamed A, Traeholdt C,Cheyne SM, Sunarto S, Jayasilan M-A, Ross J, Shapiro AC,Sebastian A, Dech S, Breitenmoser C, Sanderson J, Duckworth JW,Hofer H. 2010. Modelling the species distribution of flat-headedcats (Prionailurus planiceps), an endangered south-east Asiansmall felid. PLoS ONE 5:e9612.

Wonkka CL, Rogers WE, Smeins FE, Hammons JR, Ariza MC, Haller SJ.2012. Biology, ecology, and conservation of Navasota ladies-tresses (Spiranthes parksii Correll): an endangered terrestrialorchid of Texas. Native Plants Journal 13:236–243.

Wu XB, Smeins FE. 2000. Multiple-scale habitat modeling approach forrare plant conservation. Landscape and Urban Planning 51:11–28.



species distribution modelling for conservation of an

Documents